1d61147c01
ck_spinlock/mcs: Remove unnecessary fences, serialize unload.
12 years ago
79e336cfbc
ck_spinlock: Clarify ticket spinlock code.
...
Some people might be confused as far as lack of
fencing in the lock. Add a comment to clarify that
old values should not be equal to new values
of current position (where acquiring the current position
already has a global ordering).
12 years ago
27ed5f78fe
ck_spinlock: Remove unnecessary fence in DEC-based spinlock.
12 years ago
a7c4b9df65
ck_spinlock: Adopt RMO for anderson spinlock.
12 years ago
ce4cdd45c8
ck_spinlock: Adopt RMO semantics for CLH.
12 years ago
7ebbcf55b1
regressions/ck_spinlock: Disable linux_spinlock on unsupported platforms.
12 years ago
d72ca391f6
ck_bytelock: Switch to RMO model.
12 years ago
180b5e0386
ck_brlock: Serialize brlock operations for RMO.
12 years ago
7dda1edecf
ck_barrier: Serialize centralized and combining barriers.
...
Fixes regressions in PPC.
12 years ago
9e29788dde
ck_ht: Move ifdef block.
12 years ago
e5cd9fa322
ck_ht: key_length/key are only needed if PP is enabled.
12 years ago
bac1c1d26a
regressions/ck_stack: Disable various TRYMPMC benchmarks on unsupported platforms.
12 years ago
3e4aa8239a
regressions: Add _GNU_SOURCE to ck_backoff. Don't apply to all.
12 years ago
f4e50d16d8
regressions: Use _GNU_SOURCE by default on Linux.
12 years ago
4d60db247e
build/RPM: Drop release number back to 1.
12 years ago
4debf3a735
build: Bump version for next release.
12 years ago
ab310bfdb4
ck_hp: Linearize registration and recycling.
12 years ago
94475b85be
ck_epoch: Remove redundant load fence.
12 years ago
a30a89817c
ck_epoch: Follow-up to previous commit
12 years ago
acde423ce8
ck_epoch: Adopt RMO ck_pr semantics.
...
These changes primarily affect platforms with RMO semantics.
- Write out to local epoch requires atomic semantics,
adopt atomic store to epoch in ck_epoch_reclaim.
- Correct an issue where we would publish node availability
before record re-initialization.
- Execute full barrier in write_begin to serialize active flag
IFF we are not recursing into an epoch block. Serialize active
publication with respect to epoch walk.
- Use CAS-based tick operation for epoch counter. Despite higher
constant factor, we are much less likely to starve writers on
bursty workloads.
12 years ago
d37eb779cb
build: Add dist target, allow user to modify version.
12 years ago
45f310893c
ck_rwlock: Fix behavior for weakly-ordered systems.
...
As ck_pr semantics were still not molded, I was designing
under the assumption I would potentially go towards
acq/req interface. Since RMO will be the semantic norm for
the ck_pr model from now on, enforce stricter ordering
requirements on rwlock.
ck_rwlock_write_unlock function will now also serialize both
loads and stores.
12 years ago
8043f52130
ck_pr: Add ck_pr_barrier for compiler barrier.
...
CK_CC_BARRIER isn't idiomatic, roll this into PR memory model
instead.
12 years ago
1e8fe57028
ck_bitmap: Remove unnecessary whitespace.
12 years ago
e3cf015723
build: Drop dist target, it really adds little value.
...
Use git-archive instead or similar if you need to build
your own distribution tarball.
12 years ago
7e9a0d8751
regressions/ck_brlock: Reduce number of iterations.
12 years ago
d4abd39dd6
build: Remove tarball.
12 years ago
f8c19a89b4
regressions/ck_hp: LLVM's optimizer keeps getting the best of us.
...
If I see more regressions in LLVM, I may end up enforcing compiler
barriers in all inlined code.
12 years ago
bc19dccb1f
ck_cc: Add CK_CC_BARRIER for compiler barrier.
12 years ago
f567513f8c
regressions/ck_hp: Add Donner's unit test.
12 years ago
06345e7917
build: Add "dist" target for people who want to build their own distribution.
...
Based on original patch submitted by Andrew Schorr.
12 years ago
b1b19c1f74
ck_ring: Memory fence is overly pedantic for parameterized enqueue.
12 years ago
753f95fb45
Merge branch 'master' of git.concurrencykit.org:ck
12 years ago
94698736c1
ck_queue: Add appropriate load fences.
12 years ago
7530ca21b7
ck_fifo: Do not forget about node recycling.
12 years ago
2d2b49a313
ck_stack: Assume weak ordering of CAS.
...
I was actually unsure of the exact memory model
I wanted for atomic RMW operations. It was
made apparent with time that I had to adopt RMO
if I didn't want to sacrifice performance. Make
sure we can assume RMO for the stack.
12 years ago
3f6a3660eb
regressions/ck_stack: Add trypush unit tests to check target.
...
Submitted by Andrew Schorr <aschorr@telemetry-inves....>.
12 years ago
a7f51fffca
ck_hp_fifo: Serialize second CAS.
12 years ago
d454f729e0
regressions/ck_fifo: Add dependency to ck_fifo.h.
12 years ago
ccbf28f0cb
ck_fifo: Remove unused variables, serialize MPMC insertion.
12 years ago
78f00eeeb6
ck_fifo_spsc: Fix implementation.
...
Do not be overly pedantic. Fix misplaced store
barrier and make sure to atomically load stub
pointer on reader side.
12 years ago
80c56d115b
ck_ring: Do not use volatile load for parametric ck_ring_size.
12 years ago
6159940337
build: Bump version for next release.
12 years ago
fc316307c4
build: Add sysctl line for OpenBSD.
12 years ago
ad5e143bd5
ck_ring: Various improvements to readability.
...
No regressions in unit tests, so I'm dropping volatile
semantics for unnecessary volatile load/store on producer/consumer
side.
12 years ago
290add2e50
build/regressions: Add CORES option, add "check" target.
...
The check target will attempt to run the unit tests with some
sane defaults.
12 years ago
fbb88a519c
ck_bytelock: read_lock must serialize all load/store before checking owner field.
...
Thanks to Andrew Schorr for reporting the unit test failure.
12 years ago
3afa6b2b39
ck_ht: Let users define their own empty and tombstone values.
12 years ago
94c26494a6
ck_ring: Size is no longer necessary.
...
Patch submitted by Andrew Schorr.
12 years ago
1a8b3db453
ck_ht: Various performance improvements.
...
Though a new implementation is in the works, roll in
some performance improvements in the mean time.
The probe routines have been broken out into separate
reader/writer variants. These variants are much less
branch-intensive (and don't involve predict to stall
in many cases).
New implementation is attempting to deal with
interface-induced overheads.
12 years ago
Samy Al Bahra
Samy Al Bahra